WORKING PAPER NUMBER 36
F
E B R U A R Y
2004
Boom Towns
and Ghost Countries:
Geography, Agglomeration,
and Population Mobility
By Lant Pritchett
Abstract
Ghost towns dot the West of the United States. These cities boomed for a period and then, for
various reasons, fell into a process of decline and have shrunk to a small fraction of their former
population. Are there ghost countries—countries that, if there were population mobility, would
only have a very small fraction of their current population? This paper carries out four empirical
illustrations of the potential magnitude of the “ghost country” problem by showing that the
“desired population” of any given geographic region varies substantially. First, the variance of
growth rates of populations due to mobility across regions of the same country is often twice large
as the variance across all developing countries in the world. While the variance of per capita
output or income growth is much smaller. The ratio of the variance of the growth of population to
the variance of the growth of output per head across regions within countries is 4 to 14 times as
large as the same ratio across developing countries. Second, using county level data I construct
“ghost regions” of the United States —contiguous collections of counties that are the size of many
countries and have only a third the population they would have had without out-migration. Third,
I compare the historical evolution of labor force and real wages of Ireland in the nineteenth century
to the response of labor force and wages (or output per head) to negative shocks when labor
mobility is restricted. Fourth, I calculate the changes in the labor force that would restore GDP per
capita to its previous peak. All of these calculations suggest that even with thorough going
“globalization” —the free mobility of goods and capital— and complete “policy reform” —common
economic institutions and policies— there will remain substantial pressures for labor mobility.
This also implies there will be both boom towns and ghost countries.
Draft: Comments Welcome
1
2/18/2004
Boom Towns and Ghost Countries:
Geography, Agglomeration, and Population Mobility
Lant Pritchett
Kennedy School of Government
and
Center for Global Development
CGD Working Paper #36
February 18, 2004
Draft: Comments Welcome
2
2/18/2004
Boom Towns and Ghost Countries:
Geography, Agglomeration, and Population Mobility1
Is Zambia a ghost country? People moved to Zambia, and to a particular region
of Zambia, in part because you could dig a hole in the ground and extract something of
value—copper ore2. Technological changes in the world economy appear to have
permanently reduced the profitability of copper mining. There are currently around 10
million people within the imaginary lines called the borders of the nation-state of Zambia.
Imagine that Zambians were free to move anywhere in the world how many would still
be within the current borders of Zambia? If a potential ghost country is defined as a
country that has far more people that the “constrained desired” amount, then my
conjecture is that Zambia is a potential ghost—but because people are not allowed to
move across borders Zambia is not an actual ghost with declining population but is a
zombie country (not an actual ghost, but the living dead)—trapped into low and falling
(and/or stagnant) income and wages by a lack of population mobility.
The post World War II has produced a historically unique economic experiment
of more nation-states and globalization without labor mobility. In 1940 there were only
65 independent countries while almost twice that many —125 countries— have been
created in the last 60 years3. Figure 1 shows the three waves of state creation—the first
1
I would like to thank the LIEP and Development Lunch groups at Harvard for helpful comments, in
particular Dani Rodrik for the idea of migration as an issue, Ricardo Hausmann for the suggesting figure 6
and Mark Rosenzweig and Robert Lawrence for pushing on theory and interpretation. Hannah Pritchett
produced the data and figures on county population movements. Eliana Carranza assisted in the final
stages.
2
I choose the example of Zambia of the example of copper and because I grew up near the world’s largest
open pit copper mine, the Bingham mine about an hour outside of Salt Lake City Utah. Since the price of
copper has fallen the mine has changed ownership three times and the town just near the mine has been in
continuous decline.
3
Other methods and sources give different numbers, but with the same direction. Alesina, Spolare and
Wacziarg (2000) report 69 in 1920, 89 in 1950 and 192 in 1995.
Draft: Comments Welcome
3
2/18/2004
wave of de-colonized states in South and East Asia following WWII, the second wave of
states, principally in Africa, and finally the new states created from the dissolution of the
Soviet Union. These new states obviously varied enormously in size (both territory and
population), resources, location, and income levels.
Figure 1: The number of sovereign states has tripled in the last 60 years
(Number of new states per year, by region)
20
18
16
ssa
mena
eca
lac
eap
other
14
12
10
8
6
4
2
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
0
Each of these nation-states is sovereign within its own boundaries—and can
determine their own economic policies, laws, regulations, political systems and who can
cross their borders. The post-war international system encouraged economic transactions
across borders by regulating the means of payment (IMF), providing a framework for
negotiating reductions in trade barriers (GATT/WTO), and providing for capital flows
(World Bank), first directly by lending and more recently by encouraging financial
liberalization. But, unlike the first era of globalization from 1870-1914 which was an
Draft: Comments Welcome
4
2/18/2004
“age of mass migration” (Hatton and Williamson 1998)4 in the post-WWII international
system each sovereign nation-state is allowed to limit all movement of persons across its
borders at its own discretion. Moreover, there is no powerful international organization
or mechanism encouraging nation-states to adopt more “liberal” policies for labor
mobility.
Theory alone can not predict the outcome of (a) the “proliferation of sovereigns”
(Braun, Hausmann, and Pritchett 2002) and (b) “everything but labor” globalization.
Two of the “workhorse” models of growth and trade implied that if two regions had fully
integrated markets for all goods and for capital and common economic policies, the
movement of persons was not “necessary” to achieve convergence in output per capita5
and/or the equalization of factor prices wages6.
The movement of capital or goods
could, in certain simple models, with certain restrictive assumptions on parameters, act as
a complete substitute for labor movements. But cross-national incomes per capita have
not converged7 nor have factor prices.
4
There were virtually open borders between Europe and areas of recent settlement (what O’Rourke and
Williamson 2002) call the “Atlantic Economy” but there were also substantial “South-South” movements
of population—e.g. the movement of ethnic Indians to Caribbean and Eastern Africa and the movement of
Chinese around East Asia.
5
This generated the obvious empirical anomaly that while the bare bones Solow model extended crossnationally appeared to predict absolute convergence the experience was of divergence in per capita incomes
across all countries both historically (Pritchett 1997, Bourguignon and Morrison 2002) and in the post 1960
data (though not across all people in the world in the recent period in which India and China are growing
fast). In addition the bare bones Solow model would predict counter-factually large interest rates during
the process of convergence (King and Levine 1995) and would predict counter-factually large capital flows
from rich countries to poor countries (Lucas 1988). It is far beyond the scope of this footnote to describe
how the three proposed theoretical resolutions of these anomalies (“conditional convergence” within a
Solow model augmented for human capital (Barro and Sala-i-Martin 1997, Mankiw Romer Weil 1992),
“new growth” theories that predict divergence (Romer 1986, Lucas 1988), and the emphasis on
“convergence clubs” (Ben-David 1994)) have fared empirically.
6
While, as with nearly any general theorem in economics the formal conditions for the “factor price
equalization theorem” were very stringent and “unrealistic” the forces implied by the theorem—that,
compared to autarky the net factor supplies change with integration and hence should change relative prices
make a great deal of sense.
7
That is, while there plausibly has been a reduction in the global personal distribution of income (Sala-iMartin 2002) because the two largest nations India and China have done relatively, especially since a
Draft: Comments Welcome
5
2/18/2004
Suppose that a realistic feature of a mode of the an international/interregional
economy are region specific “shocks” that produce, even after all accommodating
changes in capital stocks and goods, large persistent changes in labor demand.
•
If there are such shocks and population mobility is allowed and hence
regional supply of labor is elastic in the long-run one should observe large
variability in the growth rates of populations and relatively small
variability in the interregional growth of real wages.
I use three
examples of regions that have completely integrated (or integrating)
markets for goods, capital, and labor (states/provinces/prefectures within
countries, counties within the USA, countries in the “age of mass
migration”) to show that there are substantial movements of population in
the course of convergence of wages/incomes.
•
If there are such shocks and population mobility is restricted and hence the
regional supply of labor is inelastic then and one should observe small
changes in population due to mobility and correspondingly large
variability in the growth of wages (and incomes). Across countries there is
relatively little mobility of populations and the growth rates of income and
wages are enormously variable.
The consequence of a distribution of large region-specific changes in labor
demand is that there will be regions which experience large, persistent, positive shocks to
labor demand and become boom towns. But there are also geographic regions that will
experience large, negative, persistent, shocks. Since “desired” population can fall much
transition to rapid growth in China in 1978 and India in the early 1980s (Hausmann, Pritchett, and Rodrik
2004) the dispersion of GDP per capita across nations, which are the relevant unit to ‘policies’, widened.
Draft: Comments Welcome
6
2/18/2004
faster then the actual population this will create situations in which the actual population
will vastly exceed its new “desired” level.
•
If the negative shock is large enough and population movements are
allowed these regions will become ghosts.
•
If the negative shock is large and other regions prevent labor mobility then
potential ghost countries become unrealized ghosts or zombie countries
(the living dead) as nothing, besides out-migration, can prevent an
extended and permanent downward movement in wages.
I.
Framework
To get to the phenomena of ghost countries I need to start with the question:
“what is the desired population of a given geographic region and how much does that
vary over time?” This, it turns out, is a very hard question to ask because it has been
largely ignored in the economic literature, for both theoretical and practical reasons.
Theoretically, the implication of the Solow model and nearly all of its extensions is that
the question of the “desired populations” of countries never arose because: (a) with
constant returns to scale the location of production is formally indeterminate8, (b) within
these growth models there was no “land” or “resources” or specifically geographic
features, (c) if capital was fully and rapidly mobile then any incipient differences in the
marginal product of labor could be equalized by capital mobility, (d) if knowledge of the
8
Basically, for technical reasons it is just much easier to deal with models in which the aggregate
production function has constant returns to scale—as in all generations of the Solow model. But the literal
implication of constant returns to scale is that the distribution of economic activity across space is
indeterminate—not a question that can be answered within the formal apparatus of a growth model.
Practically, this was not such a great loss since the first generation of growth theorists were interested in the
source of persistent growth of output per head in the already developed economies and since nearly all of
these countries are economically (and spatially) large and since their economies did not have a strong
resource component simply ignoring the spatial allocation of economic activity was not crucial to those
questions.
Draft: Comments Welcome
7
2/18/2004
“production function” diffused then differences in “technology” could be eliminated by
moving “ideas” not people, and (e) the models were about the evolution of aggregates
within a set of boundaries that were assumed fixed. Since population movements across
countries have been tightly restricted during the entire period in which growth economics
has been formalized and elaborated it has been a plausible assumption to “close” growth
models with the assumption of zero labor mobility and even exogenous population
growth9.
However, two recent strands of research: the economics of increasing returns to
scale and agglomeration economies and the empirical literature on the determinants of
growth, are putting geography and population mobility back on the map (so to speak).
Trade economists have re-introduced increasing returns to scale into models of
international trade (Helpman and Krugman 1985) and from there into spatial economics
(Krugman 1991)10. This combination has led to the use of the new formal apparatus for
addressing fundamental questions like “what determines the number and location of
cities?” (e.g. Fujita, Krugman, Venables 2000). A second element of this interest in
agglomeration economies is the continued interest in regional economies and the rise and
decline of populations of cities and regions within countries. Blanchard and Katz (1992)
document state level movements in employment in the USA and demonstrate
convergence in real wages but large, persistent, variability in population growth rates
across states. Recent papers have examined the role of both productivity changes and
9
Although fertility was made endogenous in some models (Barro-Becker, Galor and Weil).
The “new growth” models a la Romer (1986) introduced the possibility of increasing returns to scale to
some factors and hence re-introduced the possibility of agglomeration economies and scale economies.
Many of the implications of the scale effects of the “first generation” of growth models were clearly
counter-factual, the more recent vintage are at least arguably plausible, although (Jones 2003) points out the
question of whether “scale effects” are “bug or a feature” or growth models is still open.
10
Draft: Comments Welcome
8
2/18/2004
demand changes in simultaneously determining population sizes and real estate prices
(e.g. Glaeser and Gyourko, 2003).
The empirical literature estimating growth regressions has re-discovered11 the
notion that there are connections between the geographic characteristics of countries and
their economic performance. In particular the work of Sachs and Gallup (1998)
emphasized spatial factors like tropical location and transport costs (proxied by being
landlocked or having navigable rivers) on economic potential and economic growth. The
work of Sachs and Warner (1995) introduced the notion of a “resource curse”—that
countries richly endowed with natural resources had, on average, lower growth rates
(Auty 1999). This work on resources has been followed by theoretical work on the
causal linkages that explain these empirical regularities of geographic features on growth
with a variety of causal paths working through “institutions” or “politics” or both12.
Moreover, authors reviewing the basic facts of economic growth are again emphasizing
the important role of agglomeration economies in explaining the spatial distribution of
output (Easterly and Levine 2002).
I.A) A general set of equations for regional growth
To discuss “growth” and regional population mobility—particularly of geographic
shifts--one needs a framework within which output, factors, and labor are simultaneously
determined. At a minimum this framework needs to specify a production function, the
11
I emphasize “re-discovered” because, to be fair to the giants on whose shoulders we stand, it was never
lost on any development economist that most poor countries were in the tropics and most rich countries
were not, or that access to waterways for transport reduced transportation costs. My conjecture about the
intellectual history whereby these facts disappeared is that, in their previous incarnation, these geographic
facts were given racist and/or colonialism rationalizing explanations and hence the discrediting of the
“explanations” discredited the “facts”—but that is conjecture.
12
There is a huge and ongoing debate about the sources of these differences. Are they purely geographic
(Sachs Warner) or are they the impact of some (perhaps long past) geographically based characteristic on
“institutions” (e.g. Sokoloff Acemoglu, Johnson, Robinson) or are they the impact of resource endowments
on current politics (Isham, Pritchett, and Woolcock, 2003).
Draft: Comments Welcome
9
2/18/2004
level and evolution of “productivity” and the evolution of accumulated factors and
finally, how population is distributed.
Suppose there are i=1,…,N spatially distinct regions—which could be either
geographically arbitrary units13 or political/administrative sub-units of countries
(districts, counties, states, provinces, prefectures).
Output determination. Output at any point in time as a function of productivity
and factors of production. I’ll only assume that the production function is separable.
1) y i = A i f ( K i , Li )
Productivity (A). I classify elements that effect the trajectory of the long-run
equilibrium productivity of the ith region under four headings: “geography” G, the
“policy” P, the “institutions” I, and the “technology” T where each capital letter
represents both current and future values14.
2) A i* = A(G i , P i , I i , T i )
Growth models often assumed that A represented exclusively something like
“technology”—a set of blueprints for “netputs”--and assumed that this technology was
freely available or diffused rapidly across regions so that all regions (within or across
countries) shared the same A. Later models have incorporated a role for policies or
institutions in affecting A as well—what is a role for “geography” in productivity?
13
For instance, the USA is divided for geological survey and cartographic purposes into “sections.”
That is, as a matter of description at this stage just think of these are multi-dimensional lists of all
possible factors that affect productivity and that whatever it is that anyone proposes (from existence of
malaria to rule of law to usable “blueprints” for production) be classified under one of those four headings.
14
Draft: Comments Welcome
10
2/18/2004
• The depletion of a location specific natural resource (mine, well, forest, soil
quality, water table) that leads to a fall in the physical productivity of factors
applied. Ghost towns around depleted mines are the classic examples15.
• A permanent fall in the relative price of a good produced with a location
specific natural resource. The impact on coffee growing regions of the advent
of new suppliers might be an (ongoing) classic example.
• Technologically induced changes increases in the physical productivity of given
geographic region or resource. Changes in agriculture technology will
differential change the productivities of lands endowed with the right
combination of soil, water, and temperature. The differential impact of the
Green Revolution strains of rice and wheat are the classic example.
Production function. A general production function could exhibit constant returns to
scale in all factors, or increasing returns in some or all factors, or external economies,
whatever. If the production function has constant returns to scale then it is impossible to
discuss formally the spatial distribution of factors within any regional space that shares all
other determinants of income in common. Increasing returns to scale or external
economies of scale are able to “predict” the clustering of production into cities but the
additional complications in modeling are considerable. The spatial equilibrium across
regions depends on the balance of centrifugal and centripedal forces (Krugman 1998).
Changes in the production function can make the forces stronger or weaker. The
pressures for agglomeration are increased by either an increase in the returns to
15
For me, the origin of some of this thinking is that I grew up in a town, Idaho City, which was once a
thriving frontier town (the largest in Idaho territory) and has a population in 2000 of 458 people. Why?
Simple. There used to be gold in the river nearby and now there isn’t any commercially exploitable gold.
Draft: Comments Welcome
11
2/18/2004
scale/external economies or by shifts in the composition of demand towards the scale
intensive activity.
Transport costs. The standard “iceberg” model of transport costs assumes that the
production technology for transportation means that if one unit is shipped from the ith
region to the jth region only T ij < 1 units arrive.
3)
y j = TC i , j y i
Transport costs create a number of offsetting forces as if there were both intermediate
goods and final goods. There are trade-offs between being near the intermediate inputs
and being near the market for final demand. Krugman and Venables (1995) show that,
starting from a very high level, transport costs can cause increases in agglomeration
(population movement towards cities) and increased differences in wages across regions;
while continued falls in transport costs can cause spatial spreading of production
responding to wage differences.
Barriers to trade across regions can be modeled as increased transport costs (or
vice versa) so that “globalizing” changes can come either through reductions in transport
costs or policy changes or both.
Accumulated factors of production. The desired capital stock in a given region
depends on the returns to investors that depend on the physical productivity of additional
capital in that region (which depends on the general productivity, the production function
and transport costs) and on the appropriability of those returns. In production functions
with agglomeration effects the returns to any one investor depend on the investment
decisions of all other investors and hence raise issues of multiple equilibrium. The
appropriability of returns depends on policies (e.g. official tax rates) and institutions (e.g.
Draft: Comments Welcome
12
2/18/2004
legal and political risk). If we assume capital markets are integrated, there is a globally
available safe return r , then optimal capital stock in the ith region K i* equates the return
in the region to the global return for the equilibrium labor force.
4) r = (1 − τ ( P i , I i )) *
∂[ A(G i , P i , I i , T i ) f ( K i* , Li* )]
∂K
Given the long-run equilibrium capital stock to get to growth one needs to specify
an equation of motion—the dynamics of the adjustment of the capital stock given its
long-run value. I am agnostic about the specification—in particular because the usual
linear dynamics imply symmetry to increases and decreases in the capital stock that is
implausible for large deviations.
5) k& = g K ( K ti − K i* , AS K )
I.B) Allocation of population/labor across regions
The above formulation of output allows a definition of a region’s “constrained
desired” population. That is, one can imagine the kth person living in region i can move
into region j only by incurring some actual moving cost C
i , j ,k
. The regions differ in
attractiveness due to their income but also a set of other factors that affect utility/wellbeing (e.g. language, culture, ‘ethnicity’, temperature/climate, congestion) for a person
living in that region—Zi,k.
Also, over and above potential income and other feature that determine utility, the
de facto and de jure legal and administrative restrictions on population mobility create a
set of individual and bilateral region specific set of “virtual taxes” τ i , j ,k . I assume the
virtual tax on remaining in the region of one’s birth is zero. These are “virtual” taxes
Draft: Comments Welcome
13
2/18/2004
because they are expressed as the price equivalent of what might be binding quantity
constraints—that is since borders are enforced using the compulsion of police/military
power the virtual tax could be large enough to prevent all movement.
This “virtual tax” formulation of the constraints on population mobility can
encompass a range of alternatives16. One case is “free population mobility” with zero
virtual taxes across all pairs of regions and for all persons. In this case utilities are
equalized across all regions (not necessarily per capita incomes, even in real terms,
because of the “Z”).
Another paradigm case is “full national mobility/zero international mobility.” Lets
assume that country C includes J C regions (and the “rest of world” is J − C ) and is
represented as with zero internal virtual taxes within the country ( τ i , j ,k = 0, ∀i, j ∈ J C ). If
person k is born in the boundaries of C, and a virtual tax is infinity (as it can be with
coercion) τ i , j ,k = ∞, ∀ j ∈ J − C for all regions not in the country —both for people
moving from C to other regions or from other regions to C—, then the non-virtual tax
inclusive utility is not being equalized across countries and hence negative geographic
shocks can produce widening differences in wages/incomes/utility across regions.
Of course, between these two extremes of “full national/no international” and
“full global” labor movement there are many possible variants. Some countries may not
allow full internal mobility (e.g. China). In other places there may be full mobility across
some countries and not others (e.g. the EU). Finally, even if labor mobility is banned or
limited de jure, the de facto situation conditional on current and anticipated enforcement
of labor laws creates a very different set of virtual taxes.
16
This draws heavily on the Neary and Roberts (1980) formulation of “virtual prices” as a modeling
technique for quantity constraints on allocations and trades.
Draft: Comments Welcome
14
2/18/2004
Finally, the virtual taxes on labor mobility can be person specific as immigration
regimes of the jth region could allow “high skill” but not “low skill” immigration or
“family unification” mobility, or whatever basis the legal regime wishes to recognize.
The constrained desired population of a region is simply the population of all
individuals who prefer to live in region i given the anticipated long-run equilibrium
income of region i with its fully adjusted stocks of capital and labor and the set of virtual
taxes in all other regions17.
6) Li* =
K(J )
∑I
k
{U k ( y i* , Z i ,k , t i , j ,k ) ≥ U k ( y j* , Z j ,k , C i , j ,k ,τ i , j ,k ), ∀ j ∈ J}
k =1
The desired population is the population with zero virtual taxes.
Finally, the labor force/population needs an equation of motion for the adjustment
of actual labor to the constrained desired level:
7) l& = g L ( Lit − Li* , AS L )
This raises the obvious problem that the population within a given region depends
both on the “rate of natural increase” (the excess of births over deaths within the region)
and on the mobility of persons across regions. While many researchers have made
fertility and mortality behavior endogenous to economic growth and assumed no labor
mobility—I am going to do the opposite and assume, for now, that the rate of natural
increase is exogenous in order to focus on mobility.
I.C) How much variability in output, how much variability in labor
17
An enormous problem with this approach is how exactly to make fertility endogenous. If we take K(J) as
all people now alive then the “constrained desired” population holds population fixed—but over long
periods one fertility is dependent on opportunities for mobility and anticipated income prospects.
Draft: Comments Welcome
15
2/18/2004
How important are spatially specific (geographic and/or agglomeration) factors
in determining growth rates of incomes and populations across regions? The above
framework leads to a set of equations that specify the growth rates of per capita income
and the labor force as jointly endogenous variables that are potentially driven by a whole
host of determinants —among which are at least two that are spatially specific:
geographic specific elements of productivity and agglomeration economies in factor
accumulation.
y& ti,t − n = y (G, P, I , T , TC , f , Z , AS K , L ,τ i , − i )
l&i = l (G, P, I , T , TC , f , Z , AS K , L , Z ,τ i , − i )
t ,t − n
Each of the variables (e.g. G, P) represents an entire trajectory as observed growth rates
over any given period t-n to t can be caused by changes in the past leading to
adjustments, anticipated changes in the future or even steady state differences in
trajectories that depend on levels.
Theory usually assumes away geographic factors for three good reasons—if one
is modeling a large advanced industrial country like the USA, Japan or France. First, in a
spatially large economy that encompasses a large number of distinct regions the
likelihood of any one, or even a set of, geographic factors affecting aggregate growth is
small. Second, for advanced economies the share of directly natural resource based
industries (agriculture, mining, fisheries) is small. Third, fundamental geographic factors
that affect levels and whose impact on output has not changed recently (e.g. climate)
could be assumed away in modeling growth.
But none of these three are true in general of the non-OECD countries. First,
dividing geographic space up into smaller and smaller units this increases the potential
Draft: Comments Welcome
16
2/18/2004
for geographic, policy, or institutionally induced variability in the income fundamentals
(G,P,I,T). This is not to say that “small is bad”—as the “boom towns” like Hong Kong
and Singapore are obviously very small. Second, the exports are often both natural
resource based and concentrated in many non-OECD countries. Third, there are changes
to spatial factors that strongly affect output.
Labor mobility is not important if desired populations do not change. Desired
populations might not change (much) if either (a) the fundamentals don’t change or (b)
the mobility of other factors can compensate. The attractiveness of the regions might not
change because there are no changes in the interaction of geography and economics that
cause people for first want to be in one place and then not want to be there. So, labor
mobility is not important for Antarctica because no substantial populations ever moved
there—its attractiveness for human populations has not changed. But the classic counterexample is the exploitation of natural resources—first people do want to be there and
then, when the extraction of the resource loses value people want to leave. The existence
of “ghost towns”—places that were once booming and attracting migration which
subsequently decline and even disappear—suggest that there is variability to desired
populations.
But even if there are regional shocks there might not be large variations in desired
population if the mobility of other factors can compensate. So, suppose a region attracts
population because it relies on one type of economic activity and then some natural or
economic shock makes that activity no longer viable. There is no longer any reason for
people to be there as opposed to any other place—but they are there. One possibility is
that new activities are created and resources (capital) flows to that place and people
Draft: Comments Welcome
17
2/18/2004
sustain roughly their same living standards but change their activities. Certainly, in the
story of many of the major cities of the world the original reason for the cities location
has long since ceased to be relevant (e.g. fortification, transport linkages) but the city
continues to thrive.
But there are two other possibilities. One is that new resources do not flow and
people leave. The other possibility is that the desired population falls, perhaps
dramatically, but that people are not allowed to leave and hence all of the adjustment to
the variability in the desired population of regions is forced onto real wages and living
standards.
I do not have, at this stage, any particular compelling way of isolating the
importance of spatial shocks, but I do have two empirical strategies, each of which has
huge difficulties. The first strategy is to examine the variability of growth of income per
person and of population across regions within countries. This is interesting because it is
the equivalent of “complete integration”—regions of countries have similar (if not
identical) “policies” and “institutions” and have access to the same “technologies” and
there are no barriers to goods or capital movements. In these circumstances, if there are
large population movements across regions and hence large variability of population, this
is strongly suggestive of the importance of spatial factors that determine desired
population. The problem is that changes in the desirability of location could be on the
“demand” side—as people prefer living in warmer places, for instance.
The second strategy is to show that the variability of population growth is small
across countries but that the variability of growth of income per head is enormous both
across countries and within countries over time. This is consistent with large spatially
Draft: Comments Welcome
18
2/18/2004
specific shocks that cannot be accommodated by population movements and hence cause
falls in income per head. The problem is that across countries there are so many other
determinants of income that are not constant. There could be “policy” ghosts or
“institutional” ghosts—and certainly a great deal of the variation in growth rates is not
due to geographic or agglomeration.
Before moving to the empirical results I want to stress that my discussion on
“desired” population of regions has nothing to with the usual “population pessimism” that
rapidly growing populations are per se a problem or that rapid population growth is are a
significant independent cause of slower growth in output per capita. In many ways this
type of analysis gets it exactly backwards—it is not that populations are bad for output
growth it is that slow output growth is bad for the living standards of populations. That
is, the real “population crisis” is not when populations grow at 2%--all of the successful
East Asian countries began their rapid growth with high rates of population growth. The
problem is not that population is rising towards the desired level—the problem is when
desired population collapses to (or below) the actual population because of either
exogenous shocks, technological changes, or bad governments (whether population
continues to grow or not).
II.
How much change in population, how much change in income per capita?
This section proceeds in three steps, mainly with graphs, followed by a summary
table. First, in spatially large and populous countries with perfectly integrated markets
for goods and capital and (roughly) equal institutions and policies the variability across
regions of the growth rates of populations is as large, and often much larger, than the
variability of the growth of income per person. Second, variability in the growth of
Draft: Comments Welcome
19
2/18/2004
income per head is enormously larger than either population growth rates, or particularly,
than growth rates due to population mobility (the difference between actual population
growth and the rate of natural increase) across countries in which populations cannot
move freely. Third, we compare the variability of growth rates of population and income
per person across countries to that across regions of countries. The within country
variability of growth of income per person is much smaller than across countries while
the variability of population growth rates is larger—and the mobility induced variability
of populations enormously larger.
II.A) Variability of growth rates of income per person and population within countries
Barro and Sala-I-Martin (1997) have shown that regions within countries show a
strong tendency for convergence (both absolutely and conditionally). I use the data from
their book on income per head and population within regions of countries (which could
be states/provinces/prefectures/regions) to show the absolute variability of the growth
rates of the two series.
With no legal constraints to population mobility, workers/households will move
—at least in part— in response to economic opportunities. Within regions of a fully
integrated region or a larger unit, the variability in the rate of growth of output per worker
should be very small—and there should be powerful forces towards convergence in real
incomes. But this could or could not involve population movements, as a change in the
desired population of a region is the combination of shocks and that shock not being fully
accommodated by movements in other factors such as capital or by trade. Theory alone
provides no unambiguous prediction about the magnitude of inter-regional population
movements.
Draft: Comments Welcome
20
2/18/2004
Seeing is believing, a picture is worth a thousand words, etc. But one problem
with the typical comparison of growth rates of income per capita and population is that
nearly all standard software packages that display graphs scale the axes independently so
it is impossible to visually infer the respective variances. So here I do two things which
are non-standard. One, in all of the figures below I scale the axes so that the range on the
vertical (growth of output per capita) and horizontal (growth of population) axes are
equal. Second, I display percentile boxes which show the vertical/horizontal distance
between the 10th and 90th percentiles18.
Figures 2 and 3 show three features of the growth of income per person and the
growth of population for the continental states of the USA and prefectures of Japan.
First, the dispersion of growth rates of population is larger than the dispersion of per
capita income growth. In the USA the 10/90 range for population is 2 ppa (.6 to 2.6)
versus only 1.2 ppa (1.3 to 2.5) for income growth. In Japan the 10/90 range is 1.9 ppa (.3 to 1.6) for population and only 1 ppa (4.8 to 5.8) for income. Second, the absolute
variability in population growth across these regions within countries is about the same as
variability in population growth rates across all non-OECD countries—and substantially
larger than the net population growth rates (10/90 ratio of .8 ppa). Third, the variability
of growth rates of income per head is spectacularly larger across countries than across
regions within countries—10/90 differences of 4.6 ppa across non-OECD countries and
only around 1 ppa for the regions of USA and Japan.
The similar figures for all of the other countries with data in the Barro and Sala-iMartin (1997) are in appendix 1. These typically show less within region variability—
18
These boxes are orthogonal to the axes as I am not so interested in displaying the covariances as I am
interested in comparing the differing variability in the two series.
Draft: Comments Welcome
21
2/18/2004
but in all of them the variability in population growth across regions is larger than
variability in growth of income per head.
Draft: Comments Welcome
22
2/18/2004
Figure 2: Per annum growth of income per person and population across
states of the United States.
Figure 3: Per annum growth of income per person and population across Japan
Draft: Comments Welcome
23
2/18/2004
II.B) Variability of growth rates of income per person and population across countries
With many sovereign states and constrained labor mobility shocks to desired
population should create wide variability across countries in output per person growth
rates and little variability of population. That is of course, exactly what one sees in
Figure 4. Per annum growth rates of GDP per capita are calculated as a least squares
growth rate over the longest period available for each country in the Penn World Table
6.0 data for all countries with at least 20 years of data. Per annum growth rates of
population are calculated for the same period. The range of growth rates is more than 9
percentage points—from over 6 ppa for Singapore and Taiwan to less than negative 2 for
Zaire, Niger, Angola. In contrast the range of population growth rates is from about 3.5
ppa (excluding the very high rates reported for Jordan) to about .5 ppa.
Draft: Comments Welcome
24
2/18/2004
Figure 4: Per annum growth of GDP per capita and
population
Of course population growth alone overstates the extent to which population
mobility has played a role as much of the cross-national variation is due to very different
rates of natural increase. To compare cross-national to inter-regional growth of
population—where nearly all of the differences are due to mobility—I use the rate of
population growth less the rate of natural increase (crude birth rates less crude death
rates) as a proxy. But this almost certainly understates actual population mobility as it is
very difficult to get estimates of population that do not, to a greater or lesser extent,
depend on extrapolations based on the rate of natural increase. The compromise is to
report both raw population growth rates and population growth net of natural increase
(which I will call “net population growth” in a simple but awkward use of ‘net’).
Draft: Comments Welcome
25
2/18/2004
As can be seen in figure 5 the variability of growth of GDP per capita is
enormously larger than the variability of net population growth rates. The standard
deviation of Y/P growth is 1.9 while the standard deviation of net population growth is
.4—almost five times smaller. The 10/90 percentile range for growth rates is 4.6 ppa
(from -.8 to 3.8) while the range for net population growth is only .8 ppa (-.5 to .3)—
almost six times smaller.
Draft: Comments Welcome
26
2/18/2004
Figure 5: Per annum growth of GDP per capita and net population (population less
rate of natural increase)
II.C) Comparing across country versus within country variability of growth rates
The variance in growth rates of populations across regions within countries with
no barriers to mobility indicates how much desired populations vary across geographic
regions—even when goods and capital markets are perfectly integrated and “policies and
institutions” are (more or less) the same. The question is whether that variability is
“large” both in an absolute sense and in particular whether this is large relative to the
observed variability in populations across countries. Moreover, if the variability of
population across countries for which there are formal barriers to mobility is low, and if
the variability of desired populations is large, one would expect the adjustment to happen
through larger variability in output per person.
Draft: Comments Welcome
27
2/18/2004
The first two columns of table 1 show that the variability of populations within
countries is large—often twice as large as the variability of income per head. Strikingly,
the standard deviation of the growth rates of population across non-OECD countries is
.78--than the population growth variability within regions of the USA (.92), Canada (.78),
Japan (.78) or Spain (.80) and only a little more than twice that of most European
countries. To some extent these comparisons vastly understate the difference as nearly
all of the variance in population growth rates across non-OECD countries is due to
differences in rates of natural increase, as these countries are at very different points in
the demographic transition, while nearly all of the within country regional variability in
population growth is due to mobility. The variability of population net of the rate of
natural increase is very much smaller across countries that within regions of a country.
The final two columns show whether variability in the desired populations of a
given geographic region are accommodated by population flows versus changes in
growth rates. The ratio of the variability of population to income per person growth is
between 4 and 14 times as high for regions within countries as across countries. This is
consistent with a large degree of variability in geographic specific productivity which is
no eliminated by goods or capital movements and which results in population movements
where allowed and in large variability in income per head where people are not allowed
to move.
Draft: Comments Welcome
28
2/18/2004
Table 1: Ratios of the dispersion of growth rates of population and GDP/income
per person across regions within countries and across countries
Ratio
Country/set of
countries
σ Pˆ
σY
P
Ratio of ratio of std. dev. for regions
within country to ratio of std. dev. Across
non-OECD countries
Ratio
Pˆp10 − Pˆp 90
Yˆ
P p10
ˆ
− YP p 90
Population growth
in non-OECD
countries net of rate
of natural increase
Population
growth
Non-OECD
(net)
Non-OECD
(population)
0.22
0.17
-
-
0.41
0.28
-
-
USA
2.02
1.67
9.58
5.95
Canada
2.37
2.56
14.69
9.13
Japan
1.92
1.90
10.93
6.79
UK
1.14
1.29
7.39
4.59
France
1.02
0.83
4.79
2.98
Spain
1.20
1.69
9.73
6.04
Germany
0.80
0.73
4.18
2.60
Italy
0.82
0.78
4.47
2.78
Source: Author’s calculations.
Figures 6a,b and 7a,b illustrate the same point as the table by showing the 10/90
boxes for growth of population and income per person on the same scale for regions
within countries and across non-OECD countries.
Draft: Comments Welcome
29
2/18/2004
Figures 6 a,b: Comparing variability of growth rates of output per capita and
population within regions of countries (Japan, Canada, USA) versus across nonOECD countries
Draft: Comments Welcome
30
2/18/2004
Figures 7a,b: Comparing variability of growth rates of output per capita and
population of regions of European countries (Spain, Italy, France, Germany, UK)
versus non-OECD countries
Draft: Comments Welcome
31
2/18/2004
These results are sufficiently striking that I want to emphasize they are not a
trivial consequence of the obvious fact that people can move within regions and cannot
move across national borders. There are many ways in which population mobility is
allowed and yet not produce the above results. First, it is possible that geographic
specific shocks to desired population are small. Using the within country inter-regional
results reduces the impact of location specific policy, institution, or non-geographic
technological shocks on desired populations. Second, it is possible that geographic
specific shocks to income do not result in changes in the desired population as
movements of goods or capital mitigate shocks. The finding that with unrestricted
mobility there is convergence in incomes is not surprising. The finding that the
differences in population growth across spatially large regions within a (reasonably)
homogenous policy and institutional environment in a fully integrated economy are
absolutely and relatively large is a non-trivial empirical finding.
III.
Ghost regions of the United States
The previous section shows large differences in population growth across states of
the USA. State level results perhaps understate the degree of spatial population mobility
by smoothing over large areas—some of which increased while other decreased. Using
data across the roughly 3,000 counties in the USA illustrates the enormous degree of
spatial variation in desired populations. At the county level there are enormous changes
in populations. While population in the United States overall has doubled since 1930
there are counties that have been essentially depopulated over the sixty years from 1930
to 1990. Slope county North Dakota has seen its population fall from 4,150 to only 907,
Smith county Kansas from 13,545 to only 5078, Huerfano county Colorado from 17,062
Draft: Comments Welcome
32
2/18/2004
to only 6,009 and McDowell county West Virginia from 90,479 to only 35,233. In
contrast, there are of course counties with explosive growth in population.
These population reductions are not simply isolated cases. I assemble collections
of counties with the largest percent reduction in population from 1930 to 1990 of various
sizes. The 612 counties with the most rapid population reduction —that in 1930 added
up to 10 percent of the US population (around 12 million people)— lost 40 percent of
their 1930 population and are only 30 percent as large as they would have been had they
not experienced out-migration. The land area covered by these counties is the size of
Bolivia. The 902 counties that contained 20 percent of the US population in 1930 (24
million people) lost 27 percent of their population over the next 60 years and are only 36
percent of the counter-factual of no out-migration. The spatial area covered by these 902
counties is larger than all but about 10 countries in the world.
Draft: Comments Welcome
33
2/18/2004
Table 2: Changes in populations of agglomerations of counties in the United States.
Agglomeration
of highest
Number of
percent
reduction in
counties
population
(out of 3065)
that add up
to… in 1930
1
Million
population
84
5
Million
population
321
10 percent
of US
population
612
20 percent
of US
population
902
Population
1930
(‘000s)
Percent
change
19301990
Current
population
relative to
rate of
natural
increase
counterfactual
Total area
(square
miles)
Number of
countries
smaller in
land area
118
(Uganda,
Syria)
160
(Turkey)
1,015.1
-64.0
17.8%
89111
5,006.4
-52.2
23.6%
309358
12,140.2
-39.3
30.0%
551345
24,916.6
-27.5
35.9%
774539
176
(Bolivia,
Mali)
181
(Mexico,
Indonesia)
Source: Author’s calculations based on county population data.
Of course, adding up counties sorted by their percentage reduction maximizes the
fall and is not directly analogous to countries as these counties are not all together. A
second exercise is to assemble counties which may cut across state boundaries but which
are contiguous and that are a shape such that it is at least conceivable that, had history
been different, a plausibly shaped state (or country) could have been formed with these
boundaries. That is, while we deliberately gerrymandered the areas to include population
losing counties we did not simply “cut out” cities or make dramatic detours to include
this or exclude that county. Gerrymandering in this way I assemble five regions of the
USA which I name—Texaklahoma (Northwest Texas and Oklahoma), Heartland (parts
of Iowa, Missouri, Kansas Nebraska), Deep South (parts of Arkansas, Mississippi,
Draft: Comments Welcome
34
2/18/2004
Alabama), Coal Pennsylvania, and Great Plains North (parts of Kansas and South
Dakota).
Even with the constraint of contiguity one can assemble spatially large territories
that have seen substantial population decline. The Great Plains North is a territory larger
than Great Britain and its population has declined 28 percent from 1930 to 1990 and is
only a bit more than a third the population than had population growth been at the rate of
natural increase. The Texalohoma region is bigger than Bangladesh and is only 31
percent the size it would have been in the absence of out-migration. We use a few
counties in the coal producing region of Pennsylvania to illustrate that not all of these
declines are explained by the decline of rural and agricultural populations— but natural
resource shocks also play a role.
Draft: Comments Welcome
35
2/18/2004
Table 3: Population change in regions (contiguous collections of counties cutting across
state borders) of the United States
Region of
the United
States
(contiguous
counties)
Ratio of
current
population
Percent
Population
to counter
change in
1930
population factual at
(‘000s)
1930-1990 rate of
natural
increase
Area
(square
miles)
Texlahoma
835.8
-36.8
0.31
58403
Heartland
1482.6
-34.0
0.33
59708
Deep South
1558.2
-27.9
0.36
36284
Pennsylvania
1182.9
Coal
-27.9
0.36
2972
Great Plains
1068.0
North
-27.7
0.36
100920
Nation
101.9%
123202.6
3536278
Number of
countries
(of 192)
with
smaller
area
(with
examples)
117/ 192
(Nicaragua,
Bangladesh)
Ratio of
area per
capita
income to
national
average
92.2%
85.2%
117/ 192
96/ 192
(Jordan,
Austria, Sri
Lanka)
43/ 192
(Trinidad
and Tobago,
Mauritius)
128/ 192
(Great
Britain,
Ghana,
Ecuador)
62.6%
84.5%
85.4%
100.0%
Source: Author’s calculations with county data.
Draft: Comments Welcome
36
2/18/2004
Figure 8: Regions of the USA have seen enormous declines in population over a
long time scale…
(Dark grey—lost more than 10,000, medium grey-lost 5-10,000, light grey lost 0-5,000, white—
gained 0-10,000, striped—gained more than 10,000)
Figure 8a: Changes in county populations in the “Heartland” region of the United
States (selected counties of Iowa, Missouri, Nebraska, and Kansas)
Osceola
Lyon
Emm et
Dic kinson
W orth
W innebago
Mitchell
Howard
Floyd
C hic kasaw
Allamakee
Wn
i neshiek
K ossuth
OÕB rien
S ioux
Dawes
S ioux
K nox
Cherry
Sheridan
Thomas
B laine
Wayne
Cherokee
Loup
Garfield
Wheeler
Ida
Humboldt
Pocahontas
S ac
Cerro
Gordo
Hancock
Crawford
Carroll
B oone
Greene
B lack
H awk
Grundy
Delaw are
Dubuque
Jones
B enton
Tam a
S tory
Clayton
B uchanan
B utler
Hardin
Hamilton
W ebster
Calhoun
Fayette
Brem er
Franklin
W right
Thurs ton
Monona
M adison St anton
Jackson
Linn
Marshall
Burt
Cuming
Clinton
C edar
Morrill
Garden
B anner
Art hur
Boone
McPherson
Logan
V alley
Platt e
Greeley
Cus ter
Cheyenne
Kimball
B uena
V ista
Dixon
W oodbury
Pierc e
Antelope
Hooker
P alo
A lto
Dakot a
Rock
Grant
P lym outh
Cedar
Holt
Brown
Box Butt e
S cott s Bluff
Clay
B oyd
K ey a P aha
Howard
P erkins
Y ork
Hall
B uffalo
Guthrie
P oweshiek
Jasper
P olk
Dallas
Iowa
Johnson
S cott
P ottawattamie
K eokuk
Mahaska
Marion
W arren
Washington
Sarpy
Montgom ery
Adam s
U nion
Lucas
Cl arke
Monroe
Jefferson
Wapello
Henry
Des
Moines
Cass
Seward
Hamilton
Madison
A dair
C ass
Louisa
S aunders
M li ls
Dawson
A udobon
Muscatine
Douglas
Butler
Polk
Merrick
Lincoln
Shelby
Washington
Nanc e
S herman
Keit h
Deuel
Harrison
Dodge
Colfax
Lanc aster
Fe
r m ont
P age
Taylor
Ringgold
A ppanoose
W ayne
Decatur
Lee
V an B uren
Davis
Otoe
Frontier
Chase
Hayes
Dundy
Hitchcock
Gosper
Adams
Kearney
Phelps
S aline
Fillmore
Clay
Franklin
Harlan
Webs ter Nuckolls
Thayer
No daway
P awnee Richardson
Rawlin s
D ec at ur
Norton
Phillip s
Sm ith
Graham
Rook s
Os born e
R ep ublic
Jewe l
Washingto n
Nemaha
Ma rshall
B row n
M itchell
Riley
Ellis
Leavenw orth
Jefferson
Wabaunsee
Dic kinson
S co tt
Wic h ita
Greele y
Lan e
R us h
N es s
Mc Pherson
Marion
H odgeman
S tanto n
Grant
A nderson
Allen
St. Cla ri
Mead e
Seward
Clark
Barber
C om anche
Phelps
i gton
Crawford W ash n
Dallas
Iron
Dent
La clede
Wilso n
N eosho
Ste.
Gen evie ve
St.
Fra ncois
Gre ene
Ja sper
Web ster
Wright
C haut auqua
Mont go mery Labet te C herokee
Bollinger
Shan non
Wa yne
Newton
Stone
McDonald
Cape
Girard eau
Rey nolds
Te xas
Lawre nce
Cowley
Harper
Perry
Ma dison
Dade
C ra wford
Su mner
Ste vens
Polk
Ba rton
Elk
Morto n
M aries
Pulask i
Cedar
Frank iln
Je ffe rson
Camden
Hic kory
Kingman
Kiow a
M liler
Linn
Bourbon
Sedg wic k
Pra tt
Hask e l
C offe y
W oodso n
St. Louis
St. Louis City
Ga scon ade
Mo rgan
Ve rnon
Gr eenw ood
B ut ler
Ford
Be nton
St. Charles
Wa rren
Cole
Os age
Henry
Chas e
R eno
E dwa rds
Gray
Mo niteau
Miami
H arve y
St afford
Lincoln
Mon tgomery
Callaway
Pettis
Bates
P aw nee
Finney
Kea rny
Fra nk lin
Ra ls
Pike
Bo one
Co oper
John son
Lyon
Barto n
Ric e
Hamilto n
Lafay ette
Johnson
D ou glas
Os ag e
Marion
Audrain
Howard
Cass
Morris
Shelby
Mo nroe
Jack son
S aline
Ellsw orth
Lewis
Ran dolph
Saline
Wy andot te
Sh aw ne e
Geary
Russ e l
Ma con
Ch ariton
Carr oll
Ray
Clay
Ot taw a
Lin coln
Treg o
Gove
Loga n
W allac e
Clark
Knox
Ca d
l well
Clinton
Platte
Pottawatomie Jack son
Cla y
Ada ri
Linn
Living ston
Bucha nan
Sheridan
Thomas
Daviess
De Kalb
Do nipha n
At chison
C loud
S herman
Sc huyler Sc otland
Sullivan
Gru ndy
Ho tl
Andre w
Cheyenne
Putna m
Me rcer
Ha rrison
Ge ntry
Gage
Jefferson
W orth
Atc hison
Johnson Nemaha
Furnas
Red W illow
Christian
Stod dard
Barry
Ho well
Ta ney
Scott
Carter
Do uglas
M ississipp i
Butler
Ore gon
Ripley
O zark
New Madrid
Pem iscot
Dunk iln
Figure 8b: Changes in county populations in the “Deep South” region of the United
States (selected counties of Arkansas, Mississippi, and Alabama)
Carroll
Bent on
Boo ne
F ult on
B axte r
Izar d
Ma dis on
Washingt on
Crawfor d
Newt on
Sear cy
Sto ne
Van Bur en
Johns on
Clay
Ran dolph
M arion
La wren ce
Craighe ad
Independen ce
Clebur ne
F ranklin
Gree ne
Shar p
Jacks on
Mississippi
Poinset t
Po pe
Cro ss
Conway
Log an
Seba stian
Whit e
Fau k
l ne r
Yell
Lono ke Prairie
Pulaski
M ontgomer y
Ga rla nd
Saline
Gran t
Arkans as
Jeffers on
Pan ola
Lime stone
Madison
La wrence
Morgan
Lee Itawamba
Pontoto c
Neva da Oua chit a
Des ha
Drew
Ca h
l o un
Tallaha tc hie
Winston
Mars hall
De Ka b
l
T ow ns
F annin
Lumpkin
Pickens
H aber sh am
W hite
Step hens
Daw son
Hall
B anksF ranklin Har t
Jackson Madison
P olk
Bloun t
Mo nroe
R ab un
Union
Gilmer
Che rokee
Cullman
Etowah
Calhoun Ch ci k asaw
Murr ay
W hitfield
C hattoo ga Gord on
Ya o
l busha
Bo il var
W alk er
F loyd B artow C her okee Fo rsyth
Marion
Lincoln
Jac kson
Colber t
Fran klin
L afaye te
Co ahoma Quitman
Da de Ca toosa
Lau derda e
l
T ishomingo
Pre ntiss
Un o
in
Clar k
Dallas Clevela nd
Hempste ad
Alcorn
Tippah
Mar shall
Monr oe
P ike
Sevie r
Litt e
l Rive r
Tate
Tun ci a
Phillips
Ho t Spring
Howar d
L ee
Benton
De Soto
St. Fr ancis
Scot t
Polk
Crittend en
Woodruff
Perr y
Haralson
P aulding
Co bb
Gwin nette
Eb
l er t
Bar row
Ca
l r ke
Oglethor pe
Ocon ee
W alton
Wilkes Lincoln
D e K alb
Douglas F ulton
R ockdale
Gree neT ali aferr o
C olumbia
C layton
N ew ton Morg an
McDuffie
Henr y
F ayette
W arr en
Ch octaw
R ci hmo nd
C ow eta
Jasper Pu tn am
Car roll
Talla dega
S palding B utts
Washington
Hear
d
Hancock
Pic kens
Tusc aloosa
Glascock
Shelby
Chico t
Clay Ra ndolph
Hump hre ys
P ike
Bur ke
Ho m
l es
Jeffer son
LamarMonr oe Jones B ald win
Attala
Noxubee
W ashing ton
Winston
Tro up Meriwether
Bibb
Sh arke y
U pson
Coosa
G reene
Jenkins
Chambers
B ibb
W li kinson
Ha rris
Chilton
Scr even
Ta
l
ap
oosa
T
albot
C
r
awfor
d
Joh nson
Ha e
l
Leake
Yazoo
Ne shoba
Ke mper
Twiggs
E manuel
T aylor
Perry
Issaq uena
P each
Lee
Mu scog ee
Elmore
Laur ens
Madison
Au tauga
Bulloch
Su mte r
Tr
eutlen
E ffingh am
Bleckley
Ho uston
Ca ndler
M arion
M acon
Sc ott
C hatahoo ch ee
War ren
Ne wton Laud erd ale
Macon
S chley
Da las
Montgomer y
Hind s
E vans
Mo ntgo mery
Pulaski D od ge
Rus sell
Ma rengo
Dooly
Ran kin
Wh eeler
Br yan
S tewar t
Webster Sumter
To ombs Tattnall
C ha th am
Lo wndes
Bullock
W li cox
Clarke
Ja sper
T elfair
C risp
Smith
C hocta w
Wilcox
Quitm an
Cla b
i orne
Libe rty
Jeff
T errell L ee
Ben H lil
Simpson
Lo ng
Ba rbou r
Davis Appling
Cop a
ih
T urner
R andolph
Pike
But e
l r
Irwin
Je fe rson
Clarke
C lay Calho un D oughe rty
W ayne
C offee B acon
Wayne
McIntosh
Jone s
Wor th
Covington
Cre nsha w
Tift
Henry E arly
Monroe
L awr enceJe fe rson
Pe
i r ce
Baker
A dam s Fr anklin
Ln
i c oln
Davis
Ber rien A tkinson
Con
ecuh
Da
e
l
Co
f
ee
Glynn
Was hington
Fo rre st
Mitchell
C olquitt
Mille r
B
rantley
War e
C ook
Covington
Mar o
in
L anier
Amite
Pike
L amar
Ho uston S eminole
Pe rr y Gr eene
Esc amb a
i
Camd en
Wilkinson
Gen eva
C linch
Walthall
C ha rle ton
Gr ady Th omas Br ooks Low ndes
Decatur
Mo bile
E chols
Sun flower Leflore
Gre nada
Cla y
We bster
L ama r
Fa yette
Walke r
St . Clair
Jefferson
Montgo mer y
Calhoun
Ce
l burne
C arroll
Oktib beha Lownde s
Br adley
Mille r
Colum bia
Lafayett e
Union
Ashley
Pear l River
Stone
Ge orge
Ba d
l win
Ha rrison
Jackson
Hanco ck
Draft: Comments Welcome
37
2/18/2004
Figure 8c: The “Pennsylvania Coal” region (selected counties of eastern
Pennsylvania)
Erie
McKean
W ar ren
Wayne
F orest
Wy oming
Cameron
E lk
Venango
Mercer
Sus quehanna
Bradford
Tioga
Potter
Crawford
Lawrence
Clearfield
Butler
Luzerne
Allegheny
Northampton
Lehigh
Schuylki ll
Juniata
B lair
Cambria
Per ry
Westmoreland
Carbon
S ny der Northumberland
Mifflin
Indiana
Monroe
Columbia
Montour
Union
C entre
Ar mstrong
B eaver
Pike
Lyc oming
C linton
Jeffer son
Cl arion
Lack awanna
Sulli van
Ber ks
Dauphin Lebanon
Bucks
Hunti ngdon
Was hington
Montgomery
Cumberland
Lanc aster
Fayette
Bedford
Som ers et
Fulton
Franklin
Greene
Chester
Philadelphia
Delaware
York
Adams
Figure 8d: County population changes in the “Great Plains North” region (selected
counties of Nebraska and South Dakota)
Di vid e
Renville
Bottineau
R olette
Bur ke
Pembina
Cavali er
Tow ner
W als h
Will iams
McHenry
Mountrail
Ramsey
Pierce
Ward
Benson
Grand
For ks
N els on
McK enzie
McLean
S her d
i an
Eddy
Wells
Foster
Dunn
S teele
Gr g
i gs
Traill
Mercer
Oli ver
B li lings
Kidder
Burleigh
Golden
V alley
Stutsman
Cass
Barnes
Stark
Morton
So
l pe
Logan
Hettinger
Ransom
La Moure
Grant
Emmons
B ow man
Richland
So
i ux
A dams
Dickey
McIntosh
M cPherson
Campbell
Corson
Sargent
Marshall
B rown
Harding
P erkins
Walworth
Edm unds
Potter
Faulk
Ro berts
Day
Grant
Dewey
Butte
Spink
Codington
Ziebach
Clark
Sully
Deuel
Hamlin
Hand
Hyde
M eade
Lawrence
S tanley
Haakon
Beadle
Hughes
Buffalo
Pennington
Jones
Sanborn
B rookings
Lake
Miner
M oody
Ly man
A urora
Brule
Jackson
Custer
Jerauld
Kn
i gsbury
Hanson
Da vison
Mn
i nehaha
McCook
Mellette
Fall River
Douglas
Tripp
Shannon
Be nnett
Todd
Hutchinson
T urner
Lincoln
Ch arles Mix
Gr egory
B on
Hom me
Ke ya Pa ha
Yankton
Union
Bo yd
Clay
Da wes
Sio ux
Kn ox
Cher ry
Sherid an
Ced ar
Holt
Br own
Pier ce
Antelo pe
Gra nt
Scot ts Bluff
Hook er
Th omas
Bla ine
Lo up
Gar field
G ard en
Arthur
Mc Phers on
Boo ne
Log an
Va lley
G ree ley
She rman
Howa rd
Cust er
Kimball
Ch eyen ne
Linc oln
Daws on
Perk ins
Plat te
Co lfax
Th urst on
Hall
Bu ffalo
Merr ick
Bu rt
Cuming
Dod ge
Was hingt on
Nan ce
Ke ith
De uel
Way ne
Madis on S tant on
Whee ler
Mo rrill
Bann er
Dix on
Dako ta
Ro ck
Box But te
Polk
Yo rk
Ha milt on
Doug las
But ler
Sa unde rs
Sar py
Ca ss
Seward
Lan cast er
Ot oe
Cha se
Hay es
Front ier
G osp er Phe lps
Kearn ey
Adams
Clay
Fillmo re
Sa line
Johns on Ne ma ha
Dun dy
Draft: Comments Welcome
Hit chco ck
Red Willow
Furn as
Har lan Fran klin
W ebst er Nuck olls
38
Thay er
Jeffers on
Ga ge
Pawn ee Rich ards on
2/18/2004
Regions within the USA can be seen as a thought experiment of what would
happen in a fully “globalized” world—geographic units linked with fully integrated
markets for land, capital, goods and labor. In such a world one can expect that incomes
would converge in levels. With the exception of the Deep South, incomes of the
population losing regions are more than 85 percent of the national average—so these
regions are poorer than the national average—but not dramatically so.
But one can ask—even with fully integrated markets with goods and capital and
(roughly) equivalent policies and institutions how much variability is there in “desired
populations” within regions? The answer: “lots.” While it may be the case that
population movements were less than they would have been because capital flowed to
these regions and goods were mobile it is still the case that the population shifts within
the United States are huge. In particular, they are vastly larger than the population shifts
one sees across the often equally arbitrary boundaries of countries in the world today.
To a large extent this is a consequence of the well known phenomena of
“urbanization” as everyone knows that the USA population went from rural to urban over
this period. Glaeser and Kohlhase (2003) show a number of factors associated with the
rise and decline of county population—e.g. having resource based industries was “bad”
having sophisticated services was “good.” My point is that it is logically possible that,
as “urbanization” proceeds and populations agglomerate into cities, there are areas of
geographic space which do not contain a single “city” in equilibrium. “Boom towns”
attract enormous inflows of population. A “ghost country” might not contain, in a full
mobility equilibrium, even a single city.
Draft: Comments Welcome
39
2/18/2004
IV.
Accommodating shocks with population movements: Ireland
The nineteenth century was an “age of mass migration” (Hatton and Williamson
1998) as many of the “areas of recent settlement” had open borders with respect to
immigrant (at least of certain ethnic/national origin). It was also an era of rapidly
reductions in transport costs and moves towards freer trade in goods, open capital
markets (and massive movements in capital)—the first era of globalization. Hence this
period is an interesting example of the question: “how we would expect geographic
shocks to be accommodated in a globalizing world?”
The story of Ireland is an obvious, and dramatic, case in point. The introduction
of the potato into Ireland has been a large positive shock-as it was a much cheaper source
of calories per unit land and was well adapted to the climate. This technological
innovation allowed population to grow without reducing output per worker—and
population in 1841 was 2.5 times that of 1754. However, given the dependence of the
Irish economy on the potato, the onset of the potato blight in the 1840s was an enormous
negative shock. After the initial terrible period of the famine the economy in fact did
reasonably well by the standard measures—real unskilled wages relative to the UK never
fell. The available aggregate figures show that GDP per capita in 1870 was more than 40
percent higher than in 1820.
The obvious mechanism of adjustment was out-migration, which was substantial.
Since we are used to see trends with steady exponential increases in population the
counter-example of Ireland is stunning. Peak recorded population of Ireland was over
eight million (8,175,000) but by 1901 population had fallen to only about four and a half
million (4,459,000)—a net loss of 3.7 million people. And this against of course rates of
Draft: Comments Welcome
40
2/18/2004
natural increase that would have led to increasing population. This contraction radically
changed the relative size of countries—Canada in 1871 was less than half Ireland’s peak
while by 1991 Canada was five times as large. Ireland was larger than Belgium in 1871
and only 60 percent as large by 1911. Even today with Ireland booming economically
Irish population is only 70 percent of its peak—the same level as 200 years ago.
Figure 9: During the period of accommodating the shock of the potato famine and
its aftermath in Ireland real wages never fell—but population fell to almost half its
previous peak
Relative to 1870=1
(Index of population, real unskilled urban wages, and GDP per capita, 1870-1)
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
20
19
10
19
00
19
90
18
80
18
70
18
60
18
50
18
40
18
30
18
20
18
10
18
Year
Pop'l
Wages/UK
GDP PC
Source: Maddison (2002) for population and GDP PC. Williamson and O’Rouke (1997) for
real wages.
Ireland’s experience with adjustment to negative shocks during an era of labor
mobility can be compared to any number of countries in more recent times—I’ll use
Bolivia first since it has comparable wage data over time19. Sometime in the late 1970s
19
Although not quite, as I had to extrapolate the wage data in the mid 1980s since the raw wage data were
anomalous, likely due to difficulties in exchange rate comparisons at official exchange rates during rapid
inflation.
Draft: Comments Welcome
41
2/18/2004
Bolivia’s economic growth slowed, stopped, and reversed due in part to negative shocks
to Bolivia’s extractive industries. How did Bolivians adjust to these shocks? Population
today is almost 90 percent higher than in 1972. In contrast, real industrial wages (at
official exchange rates) have fallen from a peak of 14 percent of US levels to only 8
percent of US levels. Real GDP per worker in 1992 was only 62 percent of its peak.
Figure 10: During the period of accommodating negative shocks Bolivia displays
exactly the opposite dynamics—population has increased by over 50 percent while
real wages and GDP per person have fallen
(Index of population, real industrial wages relative to the USA, and GDP per capita, 1972=1)
Pop'l
Wages/USA wages
92
19
90
19
88
19
86
19
84
19
82
19
80
19
78
19
76
19
74
19
19
72
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
RGDPPW
Source: Penn World Tables 6.0 for output and population. Arcetona and Rama (2003) for
industrial wages.
Comparing Ireland to Bolivia highlights the obvious: that nearly all developing
countries with negative shocks have seen their populations continue to expand rapidly
while when there was freer labor mobility in the international system labor movements
accommodated negative shocks.
Draft: Comments Welcome
42
2/18/2004
V.
Are there currently ghost countries?
One should rightly hesitate to declare that any particular territory is simply incapable
of supporting its current population at acceptable standards of living. But, on the other
hand simply maintaining a fiction because it is politically convenient for the
industrialized countries is no better. I define potential “ghost” countries (which are all,
given the lack of population mobility, zombies) as countries where (a) GDP per capita
has fallen by more than 20 percent from peak to trough (where for data purposes the peak
must come before 1990 so recent ghosts are ruled out), (b) GDP per capita today remains
less than 90 percent of peak GDP. This produces a list of 33 countries.
Of this list I have no way of showing which are “geographic” ghosts and which are
not. In particular, I have no way of knowing which of these “policy and institutional”
ghosts and which are “geographic” ghosts. That is, it could be that anticipated output fell
because of disastrously bad politics or policies, which, if reversed, would cause the area
to be enormously attractive—think of the boom Cuba is going to have when Castro is
gone for instance. To document which are geographic ghosts I would have to specify and
parameterize some particular model of location which would require grappling with the
thorny issues of increasing returns to scale, etc. Instead, I will do two calculations which
are hypothetical and simply illustrate the consequences of the possibility these countries
are ghosts.
First, because output per person has fallen in all of these countries (by definition) I
ask the question: “if desired population has received as large a negative shock relative to
its peak in this country as it has in the counter-factual [see three options below] then what
is the ratio of the post shock population to the current population?” The three count-
Draft: Comments Welcome
43
2/18/2004
factual scenarios are—“What if population in country Y has fallen relative to its
population at peak GDP per capita has fallen in country Y by as much as actual
population…
•
Fell peak to trough fall in Ireland in the 19th century (53 percent)?
•
Fell between 1930 and 1990 in three regions of the USA (Deep South, Great
Plains North, Pennsylvania Coal)—(28 percent)?
•
Rose only as fast as the bottom 10th percentile of population growth in regions
of the eight OECD countries in table 1 (.01 percent per annum).
This is obviously not “proof” about the changes in desired populations of the
countries—but just a matter of exploring the implications of plausible counter-factual
scenarios. In all of these regions GDP per capita rose substantially while populations fell.
In the countries GDP per capita has fallen while populations rose. It is at least plausible
these simply represent different adjustments to similar sized shocks to geographic
specific maximal incomes—pushing the adjustment either into wages and capital stocks
or into population movement.
Second, I ask the question: if the elasticity of GDP per person with respect to
population is negative 0.4 by how much would population have to fall in order to:
•
Restore previous peak GDP per capita, or
•
Move GDP per capita to the level it would be had it grown at 2 ppa since
the peak (roughly the world average growth rate and hence just avoids
divergence)
Table 4a begins with potentially “hard core” ghosts for three reasons. First, the
decline is more likely geographic than policy or institutional. While none of these
Draft: Comments Welcome
44
2/18/2004
countries has terrific policies or institutions—they are not the Zaire’s of the world that are
“institutional ghosts.” Second, all of these countries are landlocked, which makes the
substitution into other industries more difficult. Third, they all have “small” populations
(less than 20 million) which suggests that, in a locational equilibrium with population
mobility there might not be sufficient population for even one “city” in which case the
declines in desired population might be even more dramatic than those in the table.
Since I began with Zambia, let me use Zambia to illustrate the very simple way
the five scenarios work and the results. Zambia’s GDP per capita peaked in 1964 when
its population was 3.5 million. Today, GDP per capita is only 59 percent of its peak and
population is 10.0 million. If Zambia’s population had fallen from its 1964 level by as
much as Ireland’s actual population (48 percent) then population today would be only
1.86 million—18 percent of its current level. If Zambia’s population had fallen from its
1964 level by as much as population has fallen in three of the ghost regions in the USA
(28 percent) then its population would only be 2.52 million—25 percent of its current
level. If Zambia’s population had grown at the .01 percent of the 10th percentile in
population growth regions of eight OECD countries its population would be about what it
is today, 3.52 million—but that is only 35 percent of its current level.
The two output scenarios provide similarly striking ratios. Under the simple
assumptions made population and output per person, population would have to fall to 14
percent of its current level to raise GDP per person to the level of a non-divergent trend.
This is consistent with a negative shock roughly the magnitude of Ireland’s. To raise
output per person just to its previous peak, population would have to fall to 36 percent of
their current levels.
Draft: Comments Welcome
45
2/18/2004
Table 4a: How large is the ghosthood? (potential hard core ghosts)
Country Year of
peak
GDP
per
capita
Ratio
GDPpc2000/
GDPpcpeak
Zambia
CAF
Niger
Chad
Rwanda
Bolivia
Romania
0.59
0.44
0.50
0.50
0.75
0.87
0.74
1964
1970
1963
1979
1981
1978
1986
Ratios of the population would be to the current actual
population if…
…the labor force fell to
restore GDP per capita to
…the shock was as large as the
realized population changes in the
X assuming an elasticity
following three cases:
of output per person to
population of -.4
OECD
GDPpc
USA
Previous
Current
Ireland
lagging
implying 2
peak
population 48% fall ghost
regions
ppa growth
regions
GDPpc
from
(average
since peak
28%
0.4
1841 to
(No
fall from of p10 of
1926
Divergence)
1930population
0.4
1990
growth
(0.01%
per
annum
growth)
10089
18%
25%
35%
36%
14%
3603
27%
37%
51%
24%
11%
10832
17%
23%
32%
29%
11%
7694
30%
41%
57%
29%
17%
8508
33%
45%
63%
55%
30%
8329
33%
44%
62%
72%
34%
22435
54%
74%
103%
54%
34%
Source: Author’s calculations.
I am aware of how striking these numbers are. It is not implausible that the
desired population of the Sahel (Niger, Chad) has fallen by as much as the desired
population of the Great Plains North counties of the United States. In fact, there are
many reasons to expect the changes in desired population are probably larger in the
Sahel. If this is so then if population mobility were not constrained 3 out of every 4
people would leave Niger and this might only be enough to restore output to its level of
1963. With the simple assumed elasticities Chad just to its previous peak (1979) GDP
per capita would require 7 of every 10 people leave Chad.
Draft: Comments Welcome
46
2/18/2004
Table 4b contains a variety of other countries. Some are possible “institutional”
ghosts—which in some cases are countries in which output per person fell, at least
initially, because of wide-spread armed conflict (Mozambique, Nicaragua, El Salvador).
To some extent these are the most hopeful ghosts because it is possible that desired
population can recover very quickly. In others the fall in output is strongly associated
with deterioration in the terms of trade, particularly for oil—(e.g. Nigeria, Venezuela).
Jamaica is an interesting case—precisely because out-migration has already
played such a large role in population dynamics. By the “OECD laggard” scenario
Jamaica’s current population is 73 percent—which is because Jamaica’s population has
in fact grown so slowly, at only 1.2 ppa versus the 2.2 ppa of the rate of natural increase.
It is already the case that there are nearly as many people that are “Jamaican” living
outside of the national territory of Jamaica as within it.
Draft: Comments Welcome
47
2/18/2004
Table 4b: How large is the ghosthood?
Ratios of the population would be to the current actual population
if…
…the shock was as large as the
…the shock was as large as
realized population changes in the
the realized population
following three cases:
changes in the following
three cases:
OECD
Country Year of Ratio
Previous
GDPpc
Current
Ireland
USA
GDPpc2000/ population 48% fall
lagging
peak
peak
implying 2
ghost
regions
GDP
GDPpc
ppa growth
GDPpcpeak
from 1841 regions
(average
per
since peak
to 1926
28%
GDP
capita
(No
fall from of p10 of
Divergence)
1930population
1990
growth
(0.01%
per
annum
growth)
BDI
1977
0.60
6807
30%
40%
56%
37%
20%
TGO
1969
0.61
4527
23%
31%
43%
39%
17%
MDG
1971
0.65
15523
24%
33%
45%
43%
19%
NAM
1979
0.70
1718
31%
42%
58%
48%
25%
COG
1984
0.80
3018
33%
45%
62%
61%
36%
JAM
1972
0.80
2633
39%
53%
73%
62%
26%
GAB
1978
0.81
1230
28%
38%
53%
63%
30%
GMB
1984
0.85
1303
29%
40%
55%
69%
39%
JOR
1986
0.85
4887
30%
40%
56%
69%
42%
SEN
1960
0.89
9530
18%
24%
34%
77%
21%
COM
1968
0.59
558
24%
33%
45%
37%
16%
CIV
1979
0.61
16013
26%
35%
49%
39%
21%
CMR
1986
0.67
14876
37%
50%
69%
45%
29%
GNQ
1974
0.79
457
28%
38%
52%
60%
26%
GHA
1972
0.87
19306
25%
34%
47%
73%
29%
AGO
1973
0.39
11317
27%
37%
52%
20%
11%
MOZ
1973
0.49
17691
30%
41%
57%
28%
14%
MRT
1976
0.60
2576
29%
39%
55%
38%
19%
SLE
1970
0.62
4630
30%
41%
58%
39%
17%
SLV
1978
0.90
6276
37%
51%
71%
78%
36%
NIC
1977
0.40
5071
28%
38%
53%
21%
12%
VEN
1970
0.61
24170
24%
32%
44%
38%
17%
TZA
1987
0.64
33696
36%
50%
69%
41%
28%
PER
1975
0.86
25661
31%
43%
59%
71%
31%
ZAR
1970
0.27
46754
23%
31%
43%
13%
6%
NGA
1974
0.53
126910
25%
34%
47%
31%
16%
Source: Author’s calculations.
Draft: Comments Welcome
48
2/18/2004
Conclusion
It would be convenient in many ways if the proposition that every country can
achieve high and rising standards of living for their populations were factually true. One
reason why it would be convenient is that the present international system and
international organizations are, more or less, founded on this premise. Hence this paper
is a clear demonstration of an inconvenient fact and its implications. The inconvenient
fact is that, at least in the three cases examined—regions within countries, counties
within the USA, Ireland in the “Atlantic economy” (O’Rourke and Williamson 2002)—
when people are allowed to move they do move, in massive numbers. This strongly
suggests that even with fully integrated economies—far more integrated that
“globalization” can likely achieve—there are large changes in the “desired” populations
of areas over time20. These changes in desired population can be accommodated with
population movements and the equalization (or at least non-divergence) of incomes.
These movements create rising populations in cities and in booming areas and relative
and absolute population declines in other areas. The extreme is the creation of regions
that are a “ghost” of their former self.
Without population mobility geographic specific shocks create a different
dynamic—one of falling wages and outputs. If labor supply cannot be elastic then prices
must adjust, not quantities. The consequence is that it is plausible that even in a fully
“globalized” world and even with the best possible “policies and institutions” some
20
The apparent counter-example, which I do not examine, is the mobility of labor within the existing EU in
which the conventional wisdom is that population mobility is “lower than expected.” My suspicion is that
since this made labor mobile where it was already mobile within large countries, and for which there were
not large potential income gaps, and for economies in which geographic factors are small and
agglomeration economies are sufficient in each (there are no EU “ghost countries) that the small labor
mobility is not so unexpected.
Draft: Comments Welcome
49
2/18/2004
countries will not succeed—they are potentially ghost countries but are forced by
restrictions on population mobility into an existence as zombies.
Why do I wish to stress this inconvenient fact and the uncomfortable idea that
Zambia might be a zombie? Precisely because all current policy discussions focus only
on Zambia—the nation-state--and not on Zambians the people. International
organizations and international negotiations and international forums tend to be exactly
that—international where the actors and agents all represent nation-states and their
interests. There are structured organizations and institutions for bringing about
reductions in the barriers to the movement of goods, for the movement of finance. There
are organizations to bring about national development. But almost certainly the easiest
way to improve the living standards of a Zambian is not to improve living standards in
Zambia but to allow the person the choice to move out of Zambia. Perhaps bringing this
opportunity about needs to come more strongly on the world’s policy agenda.
Finally, in some discussions of this paper people have suggested that it was
“insensitive” to suggest even the possibility that all existing national boundaries do not
encompass a viable economy capable of sustaining their current population at anything
like a decent standard of living. But this “political correctness” critique of doubting
national viability is misguided—as it mistakes “nations” for something real. Who created
the borders? Who created the “nations”? I would argue in fact that this view has it
exactly backwards and that to insist on the interests of nation-states to control their
borders over all other considerations—including the well-being of human beings, who
through no action or fault of their own are trapped in economically non-viable regions—
is not a normatively attractive view.
Draft: Comments Welcome
50
2/18/2004
References
Acemoglu, Daron: Simon Johnson, and James Robinson (2001), "The Colonial Origins of
Comparative Development: An Empirical Investigation". American Economic Review,
December:1369-140
Auty, Richard M. (1998). “Resource Abundance and Economic Development”. World Institute
for Development Economics Research. WIDER Studies in Development Economics. Oxford
University Press. July 2001.
Barro, Robert and Xavier Sala-i-Martin (1997) “Technological Diffusion, Convergence, and
Growth". Journal of Economic Growth.
Ben-David, Dan 1994 “Convergence Clubs and Diverging Economies”. Centre for Economic
Policy Research. CEPR Discussion Paper 922.
Blanchard, Olivier and Lawrence Katz, 1992, “Regional Evolutions” Brookings Papers on
Economic Activity, 1992:1.
Bourguignon, F., and C. Morrison, (2002). “Inequality among World Citizens: 1820-1992.”
American Economic Review, September:727-44
Braun, Matias, Ricardo Hausmann, and Lant Pritchett. (May 2002). "The Proliferation of
Sovereigns: Are there Lessons for Integration." Working Paper.
Easterly, William and Ross Levine 2002 “It’s Not Factor Accumulation: Stylized Facts and
Growth Models” forthcoming: World Bank Economic Review, 15, 2001, pp. 177-219.
Fujita, M.; Paul Krugman and Anthony Venables (1999) “The Spatial Economy”. Cambridge,
MA. The MIT Press.
Glaeser, Edward and Joseph Gyourko (2003), Urban Decline and Durable Housing.
Glaeser, Edward and Janet Kohlhase (2003) “Cities, Regions and the Decline of Transport
Costs”. NBER Working Paper No. w9886
Hatton, Timothy and Jeffrey Williamson (1998) “The Age of Mass Migration – Causes and
Economic Impact”. Oxford University Press.
Helpman, Elhanan and Paul Krugman (1987) “Market Structure and Foreign Trade: Increasing
Returns, Imperfect Competition, and the International Economy”. Cambridge, MA: The MIT
Press.
Isham, Jonathan; Lant Pritchett and Michael Woolcock (2003)” The Varieties of the Resource
Experience: How Natural Resource Export Structures Affect the Political Economy of Economic
Growth”
Jones, Charles (2003) “Growth, Capital Shares, and a New Perspective on Production Functions”.
U.C. Berkeley. Mimeo.
King, Robert and Ross Levine (1992) "Financial Indicators and the Growth in a Cross Section of
Countries". World Bank Working Papers 819.
King, Robert and Ross Levine (1993) “Finance and Growth: Schumpeter might be right”.
Quarterly Journal of Economics, No. 108, pp. 717-737. MIT Press.
King, Robert and Ross Levine (1993) “Finance, Entrepreneurship, And Growth - Theory And
Evidence”. Journal of Monetary Economics 32 (3): 513-542
Draft: Comments Welcome
51
2/18/2004
King, Robert and Ross Levine (1993) "Finance, Entrepreneurship, and Growth: Theory and
Evidence". Journal of Monetary Economics, 32(3), 513-542.
King, Robert and Ross Levine (1994) "Capital Fundamentalism, Economic Development, and
Economic Growth". Carnegie-Rochester Public Policy Conference Series, Vol. 40, 994, 259-292.
Kremer, Michael (1993) "Population Growth and Technological Change: 1,000,000 B.C. to
1990," Quarterly Journal of Economics. August, 1993, pp. 681-716.
Krugman, Paul (1991) “Increasing Returns and Economic Geography,” Journal of Political
Economy.
Krugman, Paul (1991) “Geography and Trade”. MIT Press.
Krugman, Paul (1998) “The Role of Geography in Development” Paper prepared for the Annual
World Bank Conference on Development Economics, Washington, D.C.
Krugman and Anthony Venables (1995) "The Seamless World: A Spatial Model of International
Specialization," NBER Working Papers 5220.
Lucas, Robert (1988)"On the Mechanics of Economic Development," Journal of Monetary
Economics, 22(1), pp. 3-42.
Lucas, Robert (1990) Why doesn’t capital flow from rich to poor countries?” American Economic
Review 80, no. 2: 92–96.
Mankiw, Gregory; David Romer and David Weil (1992) "A Contribution to the Empirics of
Economic Growth," The Quarterly Journal of Economics, Vol. 107 (2) pp. 407-37.
Neary J. and K. Roberts (1980) “The theory of household behaviour under rationing”, European
Economic Review, 13, 25-42.
O’Rourke, Kevin and Jeffrey Williamson (2002) “Globalization and History: The Evolution of a
Nineteenth-century Atlantic Economy”. Cambridge, MA: MIT Press.
Pritchett, Lant (1997) “Divergence, Big Time”. Journal of Economic Perspectives, 11, 3–17.
Romer, Paul (1986) "Increasing Returns and Long Run Growth". Journal of Political Economy,
94, 1002-37.
Sachs, Jeffrey; Andrew Mellinger and John Gallup (1999) "Climate, Water Navigability, and
Economic Development". CID Working Paper Number 24.
Sachs, Jeffrey and Andrew Warner (2001). “The Curse of Natural Resources". European
Economic Review. Vol. 45.
Sala-i-Martin, Xavier (2002) "The Disturbing "Rise" of World Income Inequality". NBER
Working Paper 8904.
Sokoloff, Kenneth and Stanley L Engerman (1994) “Factor Endowments: Institutions, and
Differential Paths of Growth Among New World Economies: A View from Economic Historians
of the United States”. NBER Working Paper h0066.
Sokoloff, Kenneth and Stanley L Engerman (2002) “Factor Endowments, Inequality, and Paths of
Development Among New World Economics”. NBER Working Paper w9259.
Sokoloff, Kenneth and Stanley L Engerman (2003) “Institutional and Non-Institutional
Explanations of Economic Differences”. NBER Working Paper w9989.
Draft: Comments Welcome
52
2/18/2004
Appendix 1: Other regional pictures
Draft: Comments Welcome
53
2/18/2004
Draft: Comments Welcome
54
2/18/2004
Draft: Comments Welcome
55
2/18/2004
Table 5: Variance in growth of population and growth in GDP(or income) per
capita across countries and regions.
Region
Unit
Number of
units
Period
Standard
deviations of
per annum
growth rates
Pop’l
Y/P
Ratio
σP
σY
P
10/90 difference
Pop’l
Y/P
Ratio
10/90
diff.
growth
pop’l /
growth
Y/P
Recent cross-national
World
(pop’l-RNI)
World
(pop’l)
Non-OECD
(pop’l-RNI)
Non-OECD
(pop’l)
USA
Canada
Japan
United
Kingdom
France
Spain
Italy
Germany
113
122
91
99
19501990a
19501990a
.38
1.79
.21
0.8
4.4
0.18
.94
1.82
.52
.47
1.01
.47
.40
1.86
.22
0.8
4.6
0.17
.78
1.92
.41
.34
1.22
.28
1.2
1.67
0.9
2.56
1
1.90
0.7
1.29
0.6
0.83
1.3
1.69
1.1
0.73
0.9
0.78
Regions of non-European industrialized countries
States
1900.92
.45
2.02
2
48
1990
9
1926.78
.33
2.37
2.3
Provinces
1992
47
1955.84
.44
1.92
1.9
Prefectures
1990
Regions of European countries
195011
.37
.32
1.14
0.9
1990
195021
.34
.33
1.02
0.5
1990
195117
.80
.67
1.20
2.2
1990b
195020
.35
.44
.80
0.8
90
195011
.34
.39
.82
0.7
90
(a) Uses all available data to calculate growth rates, uses all countries with more than 20 years of data,
(b) Income data for Spain is for 1955-87.
Source: Based largely on regional data from Barro and Sala-i-Martin 1997.
Draft: Comments Welcome
56
2/18/2004
CENTER FOR GLOBAL DEVELOPMENT W ORKING PAPERS
No. 1, January 2002
Inequality Does Cause Underdevelopment: New Evidence,
William Easterly
No. 2, January 2002
HIV/AIDS and the Accumulation and Utilization of Human Capital
in Africa, Amar Hamoudi and Nancy Birdsall
No. 3, February 2002
External Advisors and Privatization in Transition Economies, John
Nellis
No. 4, March 2002
The Cartel of Good Intentions: Bureaucracy versus Markets in
Foreign Aid, William Easterly
No. 5, April 2002
Intellectual Property and the Availability of Pharmaceuticals in
Developing Countries, Jean O. Lanjouw
No. 6, May 2002
Winners and Losers: Assessing the distributional impacts of
privatization, Nancy Birdsall and John Nellis
No. 7, May 2002
Commodity Dependence, Trade, and Growth: When ‘Openness’ is
Not Enough, Nancy Birdsall and Amar Hamoudi.
No. 8, June 2002
Financial Crises and Poverty in Emerging Market Economies,
William Cline
No. 9, August 2002
An Identity Crisis? Testing IMF Financial Programming, William
Easterly
No. 10, Sept. 2002
Solutions when the Solution is the Problem: Arraying the Disarray
in Development, Lant Pritchett and Michael Woolcock
No. 11, October 2002
What did structural adjustment adjust? The association of policies
and growth with repeated IMF and World Bank adjustment loans,
William Easterly
No. 12, October 2002
Asymmetric Globalization: Global Markets Require Good Global
Politics, Nancy Birdsall
No. 13, October 2002
Low Investment is not the Constraint on African Development,
Shantayanan Devarajan, William Easterly, Howard Pack
No. 14, October 2002
An Index of Industrial Country Trade Policy toward Developing
Countries, William R. Cline
No. 15, October 2002
Tropics, Germs, and Crops: How Endowments Influence Economic
Development, William Easterly and Ross Levine
No. 16, October 2002
Do As I Say Not As I Do: A Critique Of G-7 Proposals On
Reforming The MDBs, Devesh Kapur
Draft: Comments Welcome
57
2/18/2004
No. 17, October 2002
Policy Selectivity Foregone: Debt and Donor Behavior in Africa,
Nancy Birdsall, Stijn Claessens and Ishac Diwan
No. 18, Nov. 2002
Private Sector Involvement in Financial Crisis Resolution:
Definition, Measurement, and Implementation, William R. Cline
No. 19, Dec. 2002
Do Rich Countries Invest Less in Poor Countries Than the Poor
Countries Themselves?, Michael A. Clemens
No. 20, December 2002 World Bank capital neither complements nor substitutes for private
capital, Michael A. Clemens
No. 21, December 2002 From Social Policy to an Open-Economy Social Contract in Latin
America, Nancy Birdsall
No. 22, January 2003
Global Economic Governance and Representation of Developing
Countries: Some Issues and the IDB Example, Nancy Birdsall
No. 23, February 2003
The Millennium Challenge Account: How much is too much, how
long is long enough?, Michael A. Clemens and Steve Radelet
No. 24, February 2003
Bootstraps Not Band-Aids: Poverty, Equity and Social Policy in
Latin America, Nancy Birdsall and Miguel Szekely
No. 25, February 2003
Privatization in Africa: What has happened? What is to be done?,
John Nellis
No. 26, March 2003
New Data, New Doubts: Revisiting “Aid, Policies, and Growth”,
William Easterly, Ross Levine, David Roodman
No. 27, May 2003
National Policies and Economic Growth: A Reappraisal, William
Easterly
No. 28, July 2003
Financing Pharmaceutical Innovation: How Much Should Poor
Countries Contribute?, William Jack and Jean O. Lanjouw
No. 29, April 2003
Economic Policy and Wage Differentials in Latin America, Jere R.
Behrman, Nancy Birdsall and Miguel Székely
No. 30, July 2003
The Surprise Party: An Analysis of US ODA Flows to Africa,
Markus P. Goldstein and Todd J. Moss
No. 31, August 2003
Privatization in Latin America, John Nellis
No. 32, September 2003 The Anarchy of Numbers: Aid, Development, and Cross-country
Empirics, David Roodman
No. 33, November 2003 Who is not Poor? Proposing A Higher International Standard for
Poverty, Lant Pritchett
No. 34, December 2003 Once more into the breach: economic growth and integration,
Andrew Warner
No. 35, January 2004
The Illusion of Sustainability, Michael Kremer and Edward Miguel
Draft: Comments Welcome
58
2/18/2004